Automatic transformation of a sequential program into a parallel form has a large
practical impact. The most common parallelization methods used are loop-level transformations based on unimodular transformations which largely use linear schedules.
In this paper, we provide a method which allows to construct multidimensional schedules for loop nests. For this, we approximate the problem of finding optimal multidimensional schedules for a loop nest by a real-valued problem which can be solved by standard algorithms for optimization problems.